Reordering DISTINCT keys to match input path's pathkeys

The ordering of DISTINCT items is semantically insignificant, so we
can reorder them as needed.  In fact, in the parser, we absorb the
sorting semantics of the sortClause as much as possible into the
distinctClause, ensuring that one clause is a prefix of the other.
This can help avoid a possible need to re-sort.

In this commit, we attempt to adjust the DISTINCT keys to match the
input path's pathkeys.  This can likewise help avoid re-sorting, or
allow us to use incremental-sort to save efforts.

For DISTINCT ON expressions, the parser already ensures that they
match the initial ORDER BY expressions.  When reordering the DISTINCT
keys, we must ensure that the resulting pathkey list matches the
initial distinctClause pathkeys.

This introduces a new GUC, enable_distinct_reordering, which allows
the optimization to be disabled if needed.

Author: Richard Guo
Reviewed-by: Andrei Lepikhov
Discussion: https://postgr.es/m/CAMbWs48dR26cCcX0f=8bja2JKQPcU64136kHk=xekHT9xschiQ@mail.gmail.com

Author: Richard Guo

doc/src/sgml/config.sgml

@@ -5360,6 +5360,20 @@ ANY <replaceable class="parameter">num_sync</replaceable> ( <replaceable class="
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-enable-distinct-reordering" xreflabel="enable_distinct_reordering">
+      <term><varname>enable_distinct_reordering</varname> (<type>boolean</type>)
+      <indexterm>
+       <primary><varname>enable_distinct_reordering</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        Enables or disables the query planner's ability to reorder DISTINCT
+        keys to match the input path's pathkeys. The default is <literal>on</literal>.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-enable-gathermerge" xreflabel="enable_gathermerge">
       <term><varname>enable_gathermerge</varname> (<type>boolean</type>)
       <indexterm>

src/backend/optimizer/path/pathkeys.c

@@ -2207,6 +2207,41 @@ pathkeys_useful_for_grouping(PlannerInfo *root, List *pathkeys)
 	return n;
 }
 
+/*
+ * pathkeys_useful_for_distinct
+ *		Count the number of pathkeys that are useful for DISTINCT or DISTINCT
+ *		ON clause.
+ *
+ * DISTINCT keys could be reordered to benefit from the given pathkey list.  As
+ * with pathkeys_useful_for_grouping, we return the number of leading keys in
+ * the list that are shared by the distinctClause pathkeys.
+ */
+static int
+pathkeys_useful_for_distinct(PlannerInfo *root, List *pathkeys)
+{
+	int			n_common_pathkeys;
+
+	/*
+	 * distinct_pathkeys may have become empty if all of the pathkeys were
+	 * determined to be redundant.  Return 0 in this case.
+	 */
+	if (root->distinct_pathkeys == NIL)
+		return 0;
+
+	/* walk the pathkeys and search for matching DISTINCT key */
+	n_common_pathkeys = 0;
+	foreach_node(PathKey, pathkey, pathkeys)
+	{
+		/* no matching DISTINCT key, we're done */
+		if (!list_member_ptr(root->distinct_pathkeys, pathkey))
+			break;
+
+		n_common_pathkeys++;
+	}
+
+	return n_common_pathkeys;
+}
+
 /*
  * pathkeys_useful_for_setop
  *		Count the number of leading common pathkeys root's 'setop_pathkeys' in
@@ -2240,6 +2275,9 @@ truncate_useless_pathkeys(PlannerInfo *root,
 	if (nuseful2 > nuseful)
 		nuseful = nuseful2;
 	nuseful2 = pathkeys_useful_for_grouping(root, pathkeys);
+	if (nuseful2 > nuseful)
+		nuseful = nuseful2;
+	nuseful2 = pathkeys_useful_for_distinct(root, pathkeys);
 	if (nuseful2 > nuseful)
 		nuseful = nuseful2;
 	nuseful2 = pathkeys_useful_for_setop(root, pathkeys);